=Paper= {{Paper |id=Vol-1260/paper2 |storemode=property |title=An Agent-based Sensor Grid to Monitor Urban Traffic |pdfUrl=https://ceur-ws.org/Vol-1260/paper2.pdf |volume=Vol-1260 |dblpUrl=https://dblp.org/rec/conf/woa/PostorinoS14 }} ==An Agent-based Sensor Grid to Monitor Urban Traffic== https://ceur-ws.org/Vol-1260/paper2.pdf

An Agent-based Sensor Grid
to Monitor Urban Traffic
Maria Nadia Postorino and Giuseppe M. L. Sarné

Abstract— The growing of vehicular traffic in urban areas has on the use of artificial neural networks (ANN) [11]–[13]. Each
worsened the citizens’ quality of life. Therefore some actions to sensor agent manages a distributed trust system in order to
reduce their negative effects and to improve transport network refine its outputs (see Section II-D and III). More in detail:
performances have been implemented over the years. To this pur-
pose, agent-based Intelligent Transport Systems can contribute (i) the ANNs process the vehicle acoustic signatures and
to manage a transport network. In this work, a non-intrusive return the traffic flow measures by limiting potential loss of
grid of agent-based sensors able to monitor traffic parameters accuracy due to environmental noise signals [14]; (ii) each
is proposed. It exploits acoustic signatures of road vehicles sensor agent which cooperates with its neighbouring sensor
and then analyses them to estimate traffic flows. Moreover, agents corrects and improves the ANNs outputs by using
cooperating neighboring agent sensors implement a trust system
to improve their performances. Some experimental results show the Trust Reputation Reliability (TRR) model [15], [16] takes
the feasibility and the advantages of the proposed solution. account of the existing interdependencies among their trust
measures (i.e. each trust measure permeates all the other trust
Index Terms—Acoustic Vehicle Signature, Multi-Agent System,
Sensors Grid, Transport System, Trust System. measures in order to obtain more reliable trust values).
A prototype of the proposed sensor agent grid has been
realized by using the agent platform JADE [17] and some
I. I NTRODUCTION tests have been performed to verify its performances. To this
Facing the increasing rate of vehicular traffic in most aim, the real data of a transport sub-network were used.
cities, government Authorities’ new strategies is to manage the In the following, Section II provides an overview of the
existing rather than to invest in new infrastructures [1]. This proposed sensor agent. The trust system is described in Sec-
current tendency is also due to environmental awareness and tion III, while the results of the real data experiments are
reduced availability of budgets for new and more expensive presented in Section IV. The Section V deals with related
investments [2], [3]. work and, finally, Section VI draws some main conclusions
In this context, a relevant aid is coming from progresses in
computer science, electronic, control systems, signal process- II. T HE S ENSOR AGENT
ing, communications and more and more sophisticated traffic This Section presents an overview about the sensor agent,
models to realize Intelligent Transport Systems (ITS) that represented in Figure 1 according to: (i) the analogic signal de-
improves transport network performances [4], [5]. Therefore, tection; (ii) the A/D signal conversion and its pre-processing;
nowadays it is easier (i) to assist drivers on their travel de- (iii) the ANN pattern analysis to return some traffic measures;
cisions by real-time traffic information systems (for instance, (iv) the traffic measure correction based on a distributed trust
directly provided on their personal devices [6]) (ii) to adopt system locally implemented by each agent.
effective traffic control strategies (e.g., restricted traffic zones;
speed limitation) starting from user travel preferences [7].
In this paper, a new approach is proposed to detect and A. Signal Detection
monitor traffic flows in order to adopt suitable traffic control Traffic detectors [18]–[20] are classified according to the
strategies. The system works in real time, requires inexpensive adopted physical principle (i.e. radio frequency, pressure,
detectors and produces very low environmental impact. More magnetic fields, audio, etc.) and their positioning (i.e. on-board
in detail, a sensor grid detects passages of vehicles and then or in/over roadway).
classifies them according to their acoustic signatures [8]. Each On board traffic detectors include the GPS-based ones [21],
sensor of the grid is associated to a software agent [9], [10], able to collect many travel data (i.e. travel time, average
an autonomous software entity coordinating all its activities speed, directions, etc.) for several transport applications [22],
and cooperating with the other sensor agents. although GPS signal could be loss, mainly due to the land
The detection of the acoustic signatures generated by mov- morphology. The in/over roadway class includes detectors
ing vehicles (see Section II-A) is based on the adoption of recognizing vehicles (and other traffic parameters) that are
simple and non-intrusive acoustic sensors, although it implies moving across a detection zone. In turn, they are classified
a complex signal processing phase that here has been based in intrusive (e.g. inductive loops and pneumatic or piezoelec-
tric tubes) and not intrusive (e.g. video, audio, infrared or
M.N. Postorino is with the Dept. DICEAM, University of Reggio Calabria, microwave detectors) [23].
Loc. Feo di Vito, 89122 Reggio Calabria, Italy, e-mail: npostorino@unirc.it
G.M.L. Sarné is with the Dept. DICEAM, University of Reggio Calabria, The first ones are subject to deterioration, while the others
Loc. Feo di Vito, 89122 Reggio Calabria, Italy, e-mail: sarne@unirc.it are susceptible to the adverse weather conditions (e.g.severe
2

2
Acoustic Sensor Filter
1

Amplitude
Analogical-Digital Conversion
0
Fragments Extraction FFT
-1

Frequency Domain -2
Features Extraction 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5

Agent Time (sec.)
Datasets / Patterns
Fig. 2. The characteristic sound produced by vehicles moving with respect
ANN 1 to a fixed point in the time domain.
ANN 2

hypothesis that about 90% of useful information belongs to the
Evaluator
range 100 ÷ 5000 KHz [12], this process1 has been performed
Reputation
System by adopting a 10 Khz sample frequency, a quantization on 16
Vehicle Class bits and a Grey codify. When vehicles are spaced for more
than 1 sec, a software procedure extracts a fragment (F ) of
1.5 seconds (centred on the peak value) from each audio track
Fig. 1. The Tasks of the sensor agent. enclosed between two gaps. According to the Doppler effect
and the used ANNs (see below), some preliminary tests shown
that a frequency spectrum analysis of such fragments allows
fog blinds video sensors) but they: (i) avoid to trouble traffic
the passage and the class of a vehicle to be identified. Then
for installation and maintenance issues; (ii) follow road di-
each audio fragment F is split into three equal slices (si , with
rection or geometry changes easily; (iii) monitor more lanes
i = 1, 2, 3) of 0.5 second (to consider the Doppler effect) and
also with a single sensor; (iv) have a low vulnerability to
converted from the time to the frequency domain with a Fast
mechanical damages. However, the choice of the best sensor to
Fourier Transformation (FFT) [29].
use relies on many factors as required data, traffic composition,
Finally, some features representing the most salient signal
road geometry, intrusiveness, installation and life, weather
characteristics have been extracted from each slice. Their type
conditions.
and number depend on the adopted analysis procedure and
In this work, non-intrusive audio detectors (microphones)
then some tests have been performed by using a trial and error
have been considered. They detect the acoustic vehicle sig-
method (see Section II-C). Consequently, each slice si was
natures generated by the interactions tire-road and by other
split in some frequency bands (fj ) from which the mean values
inside noise sources (e.g. the engine) [24], [25]. Indeed, they
of the emitted signal power has been computed to represent it.
are cheap, easy to install or remove, return several traffic
Tests identified the best balance between computational costs
data (i.e. speed, vehicle category, vehicles gap, etc.) and their
and accuracy, showed a subdivision of the frequency range
performances are quite good, although their accuracy might
100-5000 Hz in ten regions having boundary frequencies of
fail for adverse weather, stopped or very slow vehicles, high
100, 149, 220, 325, 480, 709, 1047, 1548, 2288, 3383 and
background noise level or for a wrong sensor location.
5000 Hz. Note that any useful result is possible with less of
nine frequency bands.
B. Signal Processing
Acoustic vehicle signatures main characteristic is its quick C. The ANN Component
variation in time (see Figure 2) and frequency domains for:
(i) kinematic, amount and traffic flow composition; (ii) road ANNs, inspired to the biological neural networks, fit well
geometry; (iii) weather; (iv) reflecting obstacles (e.g. buildings, the problem to recognize passage and class of a vehicle from
vehicles). Furthermore, the audio signal is apparently modified its acoustic signature [25] without having knowledge on the
because of the Doppler effect [26]. Then, it grows in intensity specific function linking input and output data. According
and frequency when the vehicle approaches the sensor (i.e. to some preliminary tests, two multilayer supervised ANNs,
microphone) and vice versa when it moves away. trained by a back-propagation (BP) algorithm [11], have been
The acoustic vehicle signature is very rich of information identified as the optimal solution in terms of architecture,
but not all of them need. To delete irrelevant information, the topology and parameters calibration.
signal processing starts with a filtering phase to cut off (i) Briefly, the BP algorithm works for patterns (examples) and
noise signals with a low intensity as, for instance, overnight modifies iteratively its learning parameters based on the error
and (ii) the frequencies over the 5 KHz (see below). between predicted and expected output values. The learning
The resulting analogical signal is converted in digital one process ends if the unknown relationship between input and
(A/D) [27], [28] by applying a Pulse Code Modulation (PCM) output is reached with the required precision. Then the trained
transformation including: (i) a sampling process to return a ANN can be directly applied to unknown patterns.
discrete-time signal with constant amplitude; (ii) a quantiza- 1 Note that a 0-20 KHz signal, 44.1 kHz sampled, 16 bit quantized and
tion on a finite number of levels; (iii) a codify. According to the codify by using the Grey code, generates more of 40.000 samples for second.
3

Specifically, we adopted three-layer ANNs with hyperbolic C. Trust
tangent and sigmoid as neuron functions for the hidden and Commonly, the trust measure that an agent a assigns to an
the output layers, respectively. Both the ANNs receive in agent b for its service (i.e. τab ∈ [0, 1] ∈ R) combines the
input 30 values for pattern, i.e. the 10 feature values extracted reliability measure ρab with the reputation measure πab . Thus
by each slice s in which is split F (see Section II-B), and the direct knowledge that a has about b and the suggestions
return real values ranging in [0, 1]. The first ANN identifies a given from the other agents to a about b are taken into account
vehicle passage, while the second one classifies it according to in the trust measure. Some approaches require to specify the
three pre-fixed categories (e.g., car, truck/bus or motorcycle). percentage of relevance given to the reliability with respect to
Therefore the first ANN has only 1 output neuron and the the reputation. In TRR τab is computed by using the parameter
second one 3. Consequently, each training pattern of the αab (i.e. αab ∈ [0, 1] ∈ R) to weight the reliability ρab and
first ANN dataset consists of 30 input and 1 output value, (1 − αab ) to weight the reputation πab . Formally, the trust
while that of the second ANN has 30 input and 3 output assigned by a to b is computed as:
values. Moreover, 3 vehicle type and noise organized on 6
different categories (e.g., rain, wind, strong wind, noise, loud τab = αab · ρab + (1 − αab ) · πab (2)
noise, background, respectively) have been considered for the Differently from the past, it is assumed that the relevance
training. of the reliability with respect to the reputation increases with
D. The Trust System Component the number of interactions iab occurred between the agents a
and b (i.e. αab = αab (iab )). In particular, αab = 1 only if iab
Each sensor agent calibrates its ANN outputs based on those is higher than or equal to a threshold N (a system parameter);
of its neighboring agents (i.e. the agents directly connected to otherwise, if αab depends on the ratio iab /N . More formally:
it on the transport network). To this aim, the sensor agent { i
exploits a distributed trust system that, according to the trust N
ab
if iab < N
αab = (3)
the agent assigns to its neighboring agents, weight the traffic 1 if iab ≥ N
values provided by them. More details about the trust system Consequently, τab can be expressed as:
are given in Section III. ∑
c∈C−{a,b} τcb · τac
III. T HE T RUST R EPUTATION R ELIABILITY M ODEL τab = αab · ρab + (1 − αab ) · ∑ (4)
c∈C−{a,b} τac
The Trust Reputation Reliability (TRR) model [15], [16],
This equation, written for all the agents, leads to a system of
is an extension - particularly a distributed version - of the
n · (n − 1) linear equations containing n · (n − 1) variables τab ,
mathematical model described in [30]. Briefly, in TRR each
where n is the number of agents. This system is equivalent to
agent has its perception of the trust (τ ) of each other agent (in
that described in [30] and admits only one solution.
its community) providing a service, for instance data based on
its reliability (ρ) and reputation (π) measures. In the following
the TRR model will be described in the detail. D. Distributed solution
When there is a wide agent community, the direct solution
A. Reliability in the TRR model of this trust model is behind the computational capabilities of
In TRR each agent a has its own reliability model inde- our sensor agent [31]. Therefore, we implemented a distributed
pendently from the other agents. Therefore, the reliability of approach where each agent applies the trust model only with
the agent b (i.e. ρab ∈ [0, 1] ∈ R) for the agent a is given by respect to its neighboring agents. In such a way, we obtain a
ρab = fa (iab ), where iab is the number of interactions that a lot of small, handling and partially overlapped trust systems,
and b performed. In other words, the level of knowledge a has were the trust values are propagated through the trust systems.
of b (i.e. iab ) due to their past interactions is considered.
IV. T HE E XPERIMENT
B. Reputation in the TRR model
This section presents the results of some experiments aimed
The agent a computes the reputation of the agent b (i.e. at verifying the effectiveness of the proposed sensor agents
πab ∈ [0, 1] ∈ R) by asking to each other agent c of its both in (i) returning number and category of the detected
community, different from a and b, an opinion about the vehicles and (ii) operating in a grid configuration on a transport
capability of b in providing a service. In TRR the opinion network. The prototypes of the agents have been realized
of c, represented by the trust measure (see below) that c has in JADE [17] and a sampling campaign has been carried
in b (i.e. τcb ), is weighted by the trust that a has in c (i.e. τac ). out in the city of Reggio Calabria, in Southern Italy. Before
Therefore, in TRR the reputation of an agent is different for describing the two experiment components, a brief overview
each agent depending on both its individual perception and on of collected data and ANN training step is presented.
the opinions of the other agents. Formally, the reputation πab The sampling campaign involved 4 detection points, for 4
is computed as the weighted mean of all the opinions (i.e. the working days and 3 sessions for day (i.e. h. 8-9, 13-14 and 18-
trust measures) of each other agent c, different from a and b, 20, when the traffic reaches its peaks) on one-ways and one-
weighted by the value of the trust that a has in c as:
∑ lane roads in different traffic and weather conditions. Each
c∈C−{a,b} τcb · τac detection point consisted of a microphone close to the road
πab = ∑ (1)
c∈C−{a,b} τac
and a notebook to store data.
4

Vehicles and Noise Recognition impact on the AN N2 performances. (iii) idle cars or bad
10
DP−A weather conditions made the sensor unable to provide useful
DP−B
8 DP−C results. More in general, tests have shown that each sensor
DP−D
DP−Av
unit overestimates slightly traffic flows and most part of the
6
Error %

errors are due to noises recognized as vehicles.
4 b) Experiment 2: According to Figure 1, the second
part of the experiment concerns the reliability of the sensor
2 grid to limit unpredictable misleading (i.e. due to temporary
0
obstacles) and the overestimation attitude of the ANN heuristic
Vehicles Noises Average
procedure. The test has been made on a small 4-detection point
real grid, see Figure 4. To this purpose, each sensor agent
Fig. 3. Traffic and Noise Recognition performed by AN N1 in recognizing computes periodically its traffic measures and sends them to
vehicles and noises at each Detection Point (DP) A-D and the average error
(DP-Av) each of its neighboring agent together with the last calculated
trust values of their common neighboring agents.
Let Fx be the traffic flow detected by the sensor agent x in
′ ′
Part of the collected data have been used to train the a time ∆t and let Fx be its weighted value computed as Fx =
ANNs (see Section II-B). Preliminary tests defined the optimal τxx · Fx , where τxx is the trust of x calculated by itself based
′
ANNs topologies (i.e. 30, 55 and 1 neurones and 30, 25 on the TRR model. Note that if Fx and Fx are greater than
and 3 neurones for the input, hidden and output layers of the maximum capability of the road (i.e. Fxmax ) then we set
AN N1 and AN N2 , respectively). In particular, the input data them to Fxmax . Moreover, let F Ix (resp. F Ox ) be the sum of
are the feature values (see Section II-B), while the output, all the incoming (i.e. ongoing) traffic flows for x provided by
′ ′
ranging in [0, 1] ∈ R, for AN N1 (resp. AN N2 ) means its neighbor agents and let F Ix ∑ (resp. F Ox ) be its weighted
′ nI ′
a vehicle passage or a noise (resp. the membership to a traffic
∑nO flow computed as F Ix = i=1 τix · F Ix,i (i.e. F Ox =
i=1 τi · F Oi ), where τi is the trust the sensor agent x has
x x x
vehicle class). The training datasets involved 2500 normalized
patterns, 50% vehicles (e.g. cars, truck/bus and motorcycles calculated for the i-th sensor about its capability of providing
with a prevalence of cars, likely to the real traffic, without trusted values. Furthermore, let Fx be the estimated traffic flow
affecting the performances [32]) and 50% noises shared on 6 of x computed as mean between its incoming and ongoing
′ ′
noise classes in equal amounts (see Section II-C). The training traffic flows above described (i.e. Fx = (F Ix + F Ox )/2) and
phases ended after about 14500 iterations for AN N1 and 9800 let δ and ψ be two learning coefficients ranging in [0, 1] ⊂ R.
for AN N2 . Note that only one ANN was unable to detect a Each sensor agent calibrates its weighted traffic measures
′
vehicle passage and/or its class with an acceptable precision. (i.e. Fx ) and reliability values (i.e. ρx ) with respect to those of
its neighboring agents on the grid by executing the following
a) Experiment 1: To verify the performances of our
heuristic procedure:
sensor agent, the trained ANNs received in input unknown
• Any correction is performed on traffic measures and
patterns as a continuous flow of acoustic signals to process ′ ′ ′

(see Sections II-B, II-C). As a result, 93.41% of AN N1 and reliability values if: (i) Fx (resp., F Ix , F Ox ) differs
88.46% of AN N2 showed an average accuracy in respectively for more than the 20% with respect to the previous step;
′

recognizing vehicle passages from noises and classifying the (ii) Fx or the i-th incoming (resp. ongoing) traffic flow
′ ′
max max
vehicles, see Figures 3 and 4. These results are interesting F Ix,i (resp. F Ox,i ) is equal to F Ix,i (resp. F Ox,i ).
′′
and close to those of the best (and more expensive) traffic • Otherwise, the final traffic flow measure of x (i.e. x ) is

detectors. However, note that: (i) some vehicles misclassi- updated as:
{ ′
fication are due to their acoustic signatures similar to that ′ F −F ′′

of other categories, for instance some vans are similar to
′′
Fx = Fx − δ · x 2 x if Fx ≤ Fxmax (5)
cars; (ii) AN N1 mistakes (i.e. noises classified as vehicles) Fxmax otherwise
and the reliability value assigned by x to each involved
sensor agent, including itself, is updated as:
Vehicle Classification { ′′ ′
F −F
12 DP−A
x ρx + ψ · xF ′ x if ρx ≤ 1
DP−B ρ = x (6)
10 DP−C
DP−D 1 otherwise
DP−Av
8
then x recomputes the trust of its neighboring agent.
Error %

6 The experiment has been performed as regards the detection
4 point C of Figure 4 by setting ∆t = 2 minutes, δ = 0.5 and
2
ψ = 0.75, while reliability and trust values of all the sensor
agents was initially set to 1. In Figure 6 the obtained results in
′ ′′
0
Car Truck/Bus Motorcycle Average terms of overestimated number of vehicles for F , F and F
with respect to the real traffic flows in the detection point C are
Fig. 4. Vehicle Classification performed by AN N2 for the AN N1 output represented. Results show that the implemented procedure is
′′
at each Detection Point (DP) A-D and the average error (DP-Av) able to obtain traffic flow measures (i.e. F ) closer to real data
5

A Since the knowledge of the traffic state on the transport
C D network is a primary need for transport planners, a large
number of sensors are currently available to detect traffic
B
data [42], as discussed in Section II-A. Some of them are
based on the analysis of acoustic vehicle signatures by ANNs,
Fig. 5. The representation of the used transport sub-network. although different pre-processing phase and ANNs are used.
In this context, in [24], [43] audio signals are processed
′ by a Linear Predictive Coding conversion, autocorrelation
than the other measures (i.e. F and F ) by taking advantage analysis and Time Delay ANNs. The authors measure traffic
of the use of a trust model in the agent grid. flows on more lane roads and classify 4 classes of vehicles,
but results are less satisfactory than those obtained in our
V. R ELATED W ORK work (although we tested only single lane roads). In [25] the
vehicle detection exploits the audio signal peaks, while the
A complete literature overview on the different aspects
classification is performed by multi-layer BP ANNs that use,
handled in this paper is beyond our aim and, therefore, only
as discriminative features, some characteristics of the emitted
those contributions coming closer to the matter presented here
acoustic energy. Authors state the classification process as
will be discussed in the following.
unreliable for vehicles different for class but similar for engine.
To monitor and manage a transport network, ITSs widely
Authors of [44] proposed to classify type and distance between
exploit the benefits provided by software agents to deal with
vehicles based on their noises for different weather and speed
large, uncertain and dynamic systems also in a distributed and
conditions. From the recorded sound signals some features
cooperative way [4], [33]–[35]. Indeed, multi-agent systems
are extracted and, after a Discrete Fourier Transformation,
are characterized by learning and adaptive capabilities when
processed by two probabilistic ANNs.
the complexity of the environments makes difficult to differ-
Finally, advances in communications, particularly wireless
ently program agent behaviors [36]. Besides, agents can take
technologies, allow wide grids to be realized [45]. Such grids
advantage of helping other agents and reciprocally share data
exploit more and more often agent technology [46] and trust
and experiences about other agents [37], as in our proposal.
systems for improving their effectiveness and performances.
Researchers are giving increasing attention to the appli-
A comparison of different trust models for grid systems is
cation of multi-agent systems to transport network control
provided in [47]. However, grids can adopt trust systems for
and management. For instance, in [33] agents cooperate to:
privacy and security reasons [48] (e.g. in presence of wireless
(i) improve the traffic management by allocating the network
sensors) and not only to identify misleading sensors [49], [50],
capacity; (ii) spread traffic information to drivers; (iii) take into
as in our case.
account drivers’ needs and preferences (in this case agents are
embedded in vehicle route assistant devices). While in [34],
[38], [39] the complex tasks involved in managing a transport VI. C ONCLUSIONS
network are decomposed into simpler agent-oriented tasks.
Agents, dynamically distributed and replaced over the trans- To monitor urban vehicular traffic, we presented an agent-
port network to adapt its management to various scenarios, based sensor using information embedded into the acoustic
are hierarchically organized on a three level architecture to: vehicle signatures to identify both passage and class of de-
coordinate agents tasks; execute the agents control; realize the tected vehicles. Moreover, this sensor agent has been designed
agents activities. to work in a grid configuration by cooperating with its
However, these management tools require to be supported neighboring agents in order to refine their measures. The
by algorithmic models able to simulate and to forecast users’ proposed sensor agent takes advantage of the adoption of
path choices [4], [40], [41] but first, and foremost, it is required neural networks for processing the audio signals and the
to know the state of the traffic on the transport network. implementation of a distributed trust system to weight the
collaboration of its neighboring sensor agents.
To test the performances of the proposed sensor agent,
Detection Point C
20
F
we built its prototype in JADE. Two different real data
F’ experiments have been realized. The first one considered the
Number of Vehicles

15 F ’’
sensor agent in a stand-alone configuration. As result, the
identification of passing vehicles and their class are close
10
to those of the best (and more expensive) traffic detectors.
The second experiment verified the effectiveness of the used
5
heuristic algorithm to refine the computed traffic measures by
exploiting a distributed trust system on a little grid of sensor
0
0 10 20 30 40 50 60 70 80 90 100 110 120 agents.
Time Intervals
Future researches will be addressed to test the performances
′ ′′
Fig. 6. Overestimated number of vehicles for F , F and F with respect to of the proposed sensor agent on both multi-lane and/or two-
the real traffic flows in the detection point C way roads and the properties of a wider grid of sensor agent.
6

ACKNOWLEDGMENT [24] A. Nooralahiyan, H. Kirby, and D. McKeown, “Vehicle classification
by acoustic signature,” Mathematical and Computer Modelling, vol. 27,
This work has been supported by the NeCS Laboratory - no. 9, pp. 205–214, 1998.
Department DICEAM - University Mediterranea of Reggio [25] J. George, A. Cyril, B. Koshy, and L. Mary, “Exploring sound signature
for vehicle detection and classification using ann,” International Journal
Calabria. on Soft Computing, vol. 4, no. 2, 2013.
[26] M. Crocker, Handbook of Acoustic. John Wileys & Sons, 1998.
[27] E. Ifeachor and B. Jervis, Digital signal processing: a practical ap-
R EFERENCES proach. Pearson Education, 2002.
[28] A. Spanias, T. Painter, and V. Atti, Audio signal processing and coding.
[1] P. Schmitt, “Managing urban change in five european urban agglomera- John Wiley & Sons, 2006.
tions: Key policy documents and institutional frameworks,” in Resilience [29] C. Burrus, Fast fourier transforms. Connexions, Rice University, 2008.
Thinking in Urban Planning. Springer, 2013, pp. 109–130. [30] F. Buccafurri, L. Palopoli, D. Rosaci and G.M.L. Sarné, “Modeling
[2] M. N. Postorino and L. Mantecchini, “A transport carbon footprint cooperation in multi-agent communities,” Cognitive systems Research,
methodology to assess airport carbon emissions,” Journal of Air Trans- vol. 5, no. 3, pp. 171–190, 2004.
port Management, vol. 37, pp. 76–86, 2014. [31] G.M.L. Sarné, “A collaborative filtering recommender exploiting a som
[3] R. Wagenvoort, C. De Nicola, and A. Kappeler, “Infrastructure finance in network,” in Recent Advances of Neural Network Models and Appli-
europe: Composition, evolution and crisis impact,” EIB papers, vol. 15, cations, ser. Smart Innovation, Systems and Technologies. Springer,
no. 1, pp. 16–39, 2010. 2014, vol. 26, pp. 215–222.
[4] B. Chen and H. Cheng, “A review of the applications of agent technology [32] S. Sampan, “Neural fuzzy techniques in vehicle acoustic signal clas-
in traffic and transportation systems,” Intelligent Transportation Systems, sification. available tohttp://202.28.199.34/multim/9733815.pdf,” Ph.D.
IEEE Trans. on, vol. 11, no. 2, pp. 485–497, 2010. dissertation, 1997.
[5] S. Li, “A survey of urban traffic coordination controls in intelligent trans- [33] V. Tomás and L. Garcia, “A cooperative multiagent system for traffic
portation systems,” in Service Operations and Logistics, and Informatics management and control,” in Proc. of the 4th Int. Joint Conf. on
(SOLI), 2012 IEEE Int. Conf. on. IEEE, 2012, pp. 177–182. Autonomous agents and multiagent systems. ACM, 2005, pp. 52–59.
[6] I. Yang and R. Jayakrishnan, “Modeling framework to analyze effect of [34] F. Wang, “Agent-based control for networked traffic management sys-
multiple traffic information service providers on traffic network perfor- tems,” Intelligent Systems, IEEE, vol. 20, no. 5, pp. 92–96, 2005.
mance,” Transportation Research Record: Journal of the Transportation [35] B. Chen, H. Cheng, and J. Palen, “Integrating mobile agent technology
Research Board, vol. 2333, no. 1, pp. 55–65, 2013. with multi-agent systems for distributed traffic detection and manage-
[7] M.N. Postorino and M. Versaci, “A neuro-fuzzy approach to simulate ment systems,” Transportation Research Part C: Emerging Technologies,
the user mode choice behaviour in a travel decision framework,” vol. 17, no. 1, pp. 1–10, 2009.
International Journal of Modelling and Simulation, vol. 28, no. 1, p. 64, [36] L. Busoniu, R. Babuska, and B. De Schutter, “A comprehensive survey
2008. of multiagent reinforcement learning,” Systems, Man, and Cybernetics
[8] P. Borkar and L. Malik, “Review on vehicular speed, density estimation (C): Appl. and Reviews, IEEE Trans., vol. 38, no. 2, pp. 156–172, 2008.
and classification using acoustic signal,” Int. Journal for Traffic and [37] S. Sen, A. Biswas, and S. Debnath, “Believing others: Pros and cons,”
Transport Engineering, vol. 3, no. 3, 2013. in Proc. of the 4th Int. Conf. on Multi-Agent Systems, ICMAS’2000.
[9] S. Su and C. Tham, “Sensorgrid for real-time traffic management,” in IEEE, 2000, pp. 279–286.
Intelligent Sensors, Sensor Networks and Information, 2007. ISSNIP [38] F. Wang and C. Wang, “Agent-based control systems for operation and
2007. 3rd Int. Conf. on. IEEE, 2007, pp. 443–448. management of intelligent network-enabled devices,” in Systems, Man
[10] B. Tierney, B. Crowley, D. Gunter, J. Lee, and M. Thompson, “A and Cybernetics, IEEE Int. Conf., vol. 5. IEEE, 2003, pp. 5028–5033.
monitoring sensor management system for grid environments,” Cluster [39] F. Wang, “Toward a revolution in transportation operations: AI for
Computing, vol. 4, no. 1, pp. 19–28, 2001. complex systems,” IEEE Intelligent Sys., vol. 23, no. 6, pp. 8–13, 2008.
[11] S. Haykin, Neural Networks - A Comprensive Foundation. New York: [40] M. N. Postorino and V. Fedele, “The analytic hierarchy process to
Macmillan College Publishing Company, 2000. evaluate the quality of service in transit systems,” in Urban Transport
[12] A. Calabró, M.N. Postorino and G.M.L. Sarné, “An acoustic passive XII. Urban Transport and the Environment in the 21st Century, 2006.
detector for traffic counts with neural networks,” in Neural Nets WIRN [41] M.N. Postorino and M. Versaci, “Modelling user mode choices by an
Vietri-01, ser. Perspectives in Neural Computing. Springer London, ellipsoidal fuzzy approach,” International Journal of Modelling and
2002, pp. 215–220. Simulation, vol. 33, no. 4, 2013.
[13] R. Jannarone, Concurrent learning and information processing: a neuro- [42] D. Middleton, D. Gopalakrishna, and M. Raman, “Advances in traffic
computing system that learns during monitoring, forecasting, and con- data collection and management,” in White papers for support workshops
trol. Springer, 2011. on Data Quality, 2003.
[14] S. Zehang, G. Bebis, and R. Miller, “On-road vehicle detection: a [43] A. Nooralahiyan, M. Dougherty, D. McKeown, and H. Kirby, “A field
review,” Pattern Analysis and Machine Intelligence, IEEE Transaction trial of acoustic signature analysis for vehicle classification,” Transporta-
on, vol. 28, no. 5, pp. 694–711, May 2006. tion Res. (C): Emerging Technologies, vol. 5, no. 3, pp. 165–177, 1997.
[15] D. Rosaci, G.M.L. Sarnè and S. Garruzzo, “TRR: An integrated [44] M. Paulraj, P. Adom, S. Sathishkumar et al., “Classification of acoustic
reliability-reputation model for agent societies,” in WOA 2011, Proc. sound signature of moving vehicle using artificial neural network.”
of the 12th, ser. CEUR Workshop Proc., vol. 741. CEUR-WS.org, Universiti Malaysia Perlis (UniMAP), 2012.
2011. [45] X. Li, J. Wu, X. Lin, Y. Li, and M. Li, “ITIS: Intelligent traffic
[16] ——, “Integrating trust measures in multiagent systems,” International information service in shanghaigrid,” in ChinaGrid Annual Conf., 2008.
Journal of Intelligent Systems, vol. 27, no. 1, pp. 1–15, 2012. ChinaGrid’08. The 3rd. IEEE, 2008, pp. 10–14.
[17] JADE URL, “http://www.jade.tilab.org,” 2012. [46] J. Gascuena and A. Fernández-Caballero, “On the use of agent technol-
[18] P. Martin, Y. Feng, X. Wang et al., “Detector technology evaluation,” ogy in intelligent, multisensory and distributed surveillance,” Knowledge
Mountain-Plains Consortium, Tech. Rep., 2003. Engineering Review, vol. 26, no. 2, pp. 191–208, 2011.
[19] M. Hallenbeck and H. Weinblatt, Equipment for collecting traffic load [47] O. Khalid, S. Khan, S. Madani, K. Hayat, M. Khan, N. Min-Allah,
data. Transportation Research Board, 2004, no. 509. J. Kolodziej, L. Wang, S. Zeadally, and D. Chen, “Comparative study
[20] D. Florea, D. Covaciu, I. Preda, and J. Timar, “Traffic counts of trust and reputation systems for wireless sensor networks,” Security
methodologies for various type of road traffic applications. available to and Communication Networks, vol. 6, no. 6, pp. 669–688, 2013.
http://aspeckt.unitbv.ro/jspui/bitstream/123456789/215/1/car20111159- [48] M. Khan, Y. Xiang, S. Horng, and H. Chen, “Trust, security, and privacy
paper.pdf,” 2011. in next-generation wireless sensor networks,” 2013.
[21] P. Misra and P. Enge, Global Positioning System: signals, measurements [49] S. Zhao, V. Lo, and C. Dickey, “Result verification and trust-based
and performance 2 Ed. Massachusetts: Ganga-Jamuna Press, 2006. scheduling in peer-to-peer grids,” in Peer-to-Peer Computing, 2005. 5th
[22] M. N. Postorino, V. Barrile, and F. Cotroneo, “Surface movement IEEE Int. Conf. on. IEEE, 2005, pp. 31–38.
ground control by means of a gps–gis system,” Journal of Air Transport [50] M. Krasniewski, P. Varadharajan, B. Rabeler, S. Bagchi, and Y. Hu,
Management, vol. 12, no. 6, pp. 375–381, 2006. “Tibfit: Trust index based fault tolerance for arbitrary data faults in
[23] E. Minge, J. Kotzenmacher, and S. Peterson, “Evaluation of non- sensor networks,” in Dependable Systems and Networks, 2005. .Proc.
intrusive technologies for traffic detection,” Minnesota Dep. of Trans- Int. Conf. on. IEEE, 2005, pp. 672–681.
portation, Research Services Section, Tech. Rep., 2010.